Developmental programming, epigenetics and obesity
During critical periods of embryonic, fetal and early postnatal development, environmental influences can affect mammalian development, leading to permanent changes in the regulation of energy balance. There is increasing acceptance of the idea that such developmental programming of energy balance regulation is an important factor in the current worldwide obesity epidemic( Reference Atkinson, Pietrobelli and Uauy 1 , Reference Gluckman, Hanson and Zimmet 2 ). To guide the exploration of underlying biologic mechanisms, 15 years ago Waterland & Garza( Reference Waterland and Garza 3 ) proposed the mechanistic construct of metabolic imprinting to describe adaptive responses to early nutrition that are characterised by a limited period of sensitivity during development (a ‘critical window’), a persistent effect lasting through adulthood, a specific and measurable outcome and a dose–response or threshold relation between exposure and outcome. Of the five potential mechanisms they elaborated( Reference Waterland and Garza 3 ), the potential for metabolic imprinting to occur via induced alterations in epigenetic gene regulation has received the most attention.
Epigenetics is the study of mitotically heritable alterations in gene expression potential that are not caused by changes in DNA sequence( Reference Jaenisch and Bird 4 ). First recognised for their importance in genomic imprinting, X-inactivation and silencing of retrotransposons( Reference Bestor, Edwards and Boulard 5 ), epigenetic mechanisms are now recognised to play a key role in stabilising gene expression potential in differentiated cells( Reference Kundaje, Meuleman and Ersnt 6 ). Various molecular mechanisms are known to participate in epigenetic regulation, including DNA methylation, covalent modification of histone proteins, autoregulatory DNA-binding proteins and non-coding RNA. Our research focuses on DNA methylation, which in mammalian somatic tissues occurs predominantly at cytosines in CpG dinucleotides. CpG methylation is recognised as the most stable epigenetic mark( Reference Cedar and Bergman 7 ), and is hence an excellent candidate mechanism of metabolic imprinting. Indeed, during critical periods of early development, nutrition and other environmental exposures can induce alterations in DNA methylation that persist to adulthood, in both rodents( Reference Dolinoy, Huang and Jirtle 8 – Reference Weaver, Cervoni and Champagne 11 ) and human subjects( Reference Dominguez-Salas, Moore and Baker 12 , Reference Waterland, Kellermayer and Laritsky 13 ).
Human and animal model data show that epigenetic dysregulation can cause obesity. As recently reviewed( Reference Waterland 14 ), adult-onset obesity in cloned mice( Reference Tamashiro, Wakayama and Akutsu 15 ) suggests epigenetic causality in obesity, and individually variable hyperphagic obesity of isogenic heterozygous agouti viable yellow (Avy ) mice, which correlates with Avy methylation( Reference Waterland and Jirtle 9 , Reference Morgan, Sutherland and Martin 16 ), provides a clear example of epigenetic dysregulation causing obesity. In human subjects, sporadic cases of Prader–Willi syndrome (a neurodevelopmental syndrome characterised by hyperphagic obesity, among other symptoms)( Reference Goldstone 17 ) are caused by aberrant hypermethylation and epigenetic silencing of a region of chromosome 15.
Despite these compelling examples, it remains unclear whether epigenetic dysregulation plays a major role in the current obesity epidemic. This is due to multiple obstacles to the study of epigenetic aetiology in human subjects, including the inherent cell-type specificity of epigenetic regulation, the influence of genetic variation on epigenetic variation, the potential for reverse causality, and the overall complexity of epigenetic regulation( Reference Waterland 14 , Reference Waterland and Michels 18 ). For these reasons, controlled studies in isogenic mouse models now offer the best opportunities to understand how interindividual epigenetic variation influences energy balance regulation. Our recent investigations into two such models have led to surprising insights.
Two mouse models of developmental programming of energy balance show alterations in physical activity but not food intake
The first of these is a mouse model of transgenerational effects of maternal obesity on the offspring. Fifteen years ago Levin proposed that in pregnancies complicated by maternal obesity the intrauterine environment may present an unfavourable milieu to the developing fetus, leading to aberrant development of central mechanisms for regulation of energy balance( Reference Levin 19 ). If, in this manner, maternal obesity during pregnancy promotes positive energy balance (i.e. energy storage in the form of adipose tissue) in her offspring, over successive generations this could lead to transgenerational amplification of obesity. Avy/a mice provide an attractive model in which to test this because they are spontaneously hyperphagic and become severely obese as adults, but unlike most genetic mouse models of obesity, the females remain fertile. We passed the obesity-promoting Avy allele through the female germline for four generations( Reference Waterland, Travisano and Tahiliani 20 ). Consistent with the feed-forward hypothesis, even among this isogenic population, the offspring of fatter mothers were themselves fatter as adults. Consequently, average adult body weight and adiposity increased over successive generations. In this experiment, we also studied another four-generation lineage that was treated identically as the first, but provided a methyl-supplemented diet known to promote DNA methylation during development( Reference Waterland and Jirtle 9 , Reference Wolff, Kodell and Moore 21 ). In the methyl-supplemented group, transgenerational amplification of obesity was completely prevented, suggesting a potential role for DNA methylation in this phenomenon( Reference Waterland, Travisano and Tahiliani 20 ). It is important to clarify that although epigenetic mechanisms affecting body-weight regulation may be involved within each generation, we believe it unlikely that the transgenerational effect in this model reflects transgenerational epigenetic inheritance. As in many examples of transgenerational effects, recapitulation is a more likely mechanism than transgenerational epigenetic inheritance( Reference Waterland 14 ).
We conducted a follow-up study in the Avy model to answer two questions: (1) When is the critical period for the effect of maternal obesity (prenatal or postnatal)?; (2) Is positive energy balance in this model due to increased food intake or reduced energy expenditure? We cross-fostered newborn mice at birth between lean a/a and obese Avy /a dams to examine the effect of being exposed to an obese Avy /a mother either during fetal development or during the suckling period( Reference Baker, Li and Kohorst 22 ). These studies provided several surprising findings. First, we found that offspring of Avy /a dams are growth restricted at birth (despite normal placental weight), suggesting placental insufficiency in Avy /a dams. When offspring of obese Avy /a dams were fostered to lean a/a dams (OL offspring) they caught up to normal body weight by weaning. Most interestingly, fetal growth restriction and subsequent catch-up growth led to increased body weight and adiposity in adulthood, but only in female offspring. Overall, although these data indicate that growth restriction during prenatal development is an essential component of the programming mechanism( Reference Baker, Li and Kohorst 22 ), they do not rule out a potential critical role for catch-up growth during the suckling period.
To answer the second question, metabolic cage studies were performed to measure home cage activity, energy expenditure and food intake. The results of these studies were completely consistent with the growth and body composition data. Whereas food intake of female OL offspring was not different from that of controls, their levels of physical activity (home cage activity) and energy expenditure were reduced directly after weaning, and remained so in adulthood. In fact, the persistently blunted spontaneous physical activity (SPA) of OL females appeared to completely explain their increased adiposity as adults( Reference Baker, Li and Kohorst 22 ). Notably, recent studies in three different human populations showed that perinatal undernutrition leads to increased obesity in adulthood, specifically in females( Reference Robinson 23 ). Our data suggest that this may be occurring by permanent blunting of physical activity.
Remarkably, in a completely different mouse model of developmental programming of obesity, we found a similar female-specific effect on physical activity. Fostering newborn rodents to small litters (SL) is a simple method to achieve overnutrition during the suckling period. When rats or mice are fostered to SL during the suckling period, by the age of weaning they are consistently heavier than control pups fostered in normal-sized litters( Reference Aubert, Suquet and Lemonnier 24 , Reference Kennedy 25 ). Suckling period overnutrition permanently alters body-weight regulation and glucose metabolism of SL rodents. Although all pups are provided the same diet (ad lib) after weaning, SL offspring remain heavier and fatter in adulthood( Reference Kennedy 25 , Reference Cryer and Jones 26 ), and exhibit impaired glucose tolerance( Reference Aubert, Suquet and Lemonnier 24 , Reference Waterland and Garza 27 ) and dysregulated endocrine pancreas function( Reference Waterland and Garza 27 ). We performed metabolic cage studies to determine if the persistent change in energy balance regulation of SL mice is caused by altered food intake or energy expenditure( Reference Li, Kohorst and Zhang 28 ). At postnatal day 1 (P1), mice were fostered to SL (four pups/litter) or control litters of nine pups. As previously reported, SL males and females were heavier and fatter than control mice at weaning (P21) and remained so as adults (P180). Our metabolic studies showed no group differences at P25, but by adulthood (P180) we observed blunted energy expenditure correlated with reduced SPA in SL mice. Although similar trends were observed in both sexes, the effects were significant (P<0·001) only in females( Reference Li, Kohorst and Zhang 28 ). Hence, together with our data in the Avy model of fetal growth restriction and catch-up growth( Reference Baker, Li and Kohorst 22 ), these findings suggest that females are particularly sensitive to developmental programming of energy balance.
Biological determinants and developmental programming of physical activity
Most importantly, integrating these two very different models of early nutritional influences on the development of energy balance regulation, in which both food intake and physical activity were extensively characterised by state-of-the-art methods, our results suggest that developmental programming of energy balance may be mediated principally at the level of physical activity. This insight is particularly timely, in that modernisation is leading to a worldwide epidemic of physical inactivity( Reference Kohl, Craig and Lambert 29 ). Physical activity clearly has a genetic component( Reference Bauman, Reis and Sallis 30 , Reference de Vilhena e Santos, Katzmarzyk and Seabra 31 ) and plays an important role in protecting some individuals from obesity( Reference Garland, Schutz and Chappell 32 – Reference Kotz, Teske and Billington 34 ). Based on a scholarly review of the literature on the biological basis of physical activity, Rowland proposed that each individual is endowed with a specific set point (the activity-stat) that determines his or her propensity for physical activity( Reference Rowland 35 ). Although most have considered the activity-stat to be genetically determined, it could also be subject to developmental programming. Indeed, in addition to our recent mouse studies, previous studies have found evidence of developmental programming of the activity stat. Before considering those studies, it is helpful to point out that physical activity is typically classified as either voluntary exercise or SPA. Voluntary exercise is physical activity that is not directly needed for survival or influenced by any outside factors, whereas SPA is all physical activity that is not voluntary exercise, including activities of daily-living and pacing and fidgeting( Reference Garland, Schutz and Chappell 32 ). In rodents, running wheel activity is the universal model for voluntary exercise( Reference Garland, Schutz and Chappell 32 ), and home cage activity is indicative of SPA( Reference Kotz, Teske and Billington 34 ).
In what we believe was the first reported animal data suggesting developmental programming of physical activity, Vickers et al.( Reference Vickers, Breier and McCarthy 36 ) studied a rat model of maternal undernutrition (30 % of ad libitum intake) throughout pregnancy, and examined effects on the offspring. They performed brief (15 min) measurements of home cage activity (SPA) when the offspring were both juveniles (P35) and adults (P145), and found reduced SPA in offspring of undernourished dams. Importantly, other than our data described above, this is the only study we are aware of that measured effects on physical activity both shortly after weaning and in adulthood. In this manner, the study demonstrated that maternal undernutrition during pregnancy caused reduced physical activity in the offspring shortly after weaning, and this persisted to adulthood. Two subsequent studies examined variants of another rat model of gestational undernutrition, assigning pregnant dams to a normal or low-protein diet during gestation( Reference Sutton, Centanni and Butler 37 ) or gestation through weaning( Reference Ohishi, Wang and Akane 38 ). In the first study, offspring of control and undernourished dams were cross-fostered at birth to control dams to limit the exposure (undernutrition) to prenatal development. Voluntary exercise (wheel running) was measured over a 2-week period in adolescent (P56) offspring of undernourished dams. Both male and female offspring of undernourished dams engaged in much less voluntary exercise than control offspring( Reference Sutton, Centanni and Butler 37 ). The other study of maternal low-protein diet( Reference Ohishi, Wang and Akane 38 ) differed from the first in that the exposure started at day 10 of gestation and continued through the end of the suckling period. Rather than reduced physical activity, this study reported increased motor activity in adult (P72) offspring of undernourished dams. Interestingly, as in our studies, this effect was observed only in female offspring( Reference Ohishi, Wang and Akane 38 ). Combining our data (discussed above) with these earlier studies, the results of these four diverse models of developmental programming suggest that, indeed, the setting of one's activity stat can be affected by nutritional exposures during prenatal and early postnatal development. Now, the challenge is to understand how this occurs, and determine whether similar processes occur in man.
Assessing the role of epigenetic mechanisms in regulation of physical activity: translation to human subjects
Accordingly, a current goal is to use these models to test the overall hypothesis that developmental programming of physical activity occurs via metabolic imprinting of epigenetic mechanisms. We began by asking two key questions. The first: What tissues are involved? Since epigenetic regulation is tissue- and cell-type specific, it is important to determine which tissues and cell types are most likely to play a primary role in the imprinting mechanism. The hypothalamus is a logical starting point because, in addition to playing a key role in regulating food intake, it is also critical in the regulation of physical activity. Peripheral factors such as cardiovascular and skeletal muscle function are important determinants of physical exercise capacity but play negligible roles in SPA. SPA appears to be primarily regulated in the hypothalamus (by key molecular players including orexins, agouti-related peptide and neuromedin U)( Reference Kotz, Teske and Billington 34 ). Voluntary exercise is also regulated in the lateral hypothalamus( Reference Rhodes, Garland and Gammie 39 ), but dopaminergic signalling (particularly in the nucleus accumbens and hippocampus) is also important( Reference Garland, Schutz and Chappell 32 ).
This brings us to the second key question: When are the critical windows during which environmental influences are most likely to affect developmental epigenetics in the hypothalamus? Our and others’ previous work indicates that epigenetic processes are most vulnerable to environmental perturbation during periods when epigenetic mechanisms are undergoing developmental establishment or maturation( Reference Waterland and Michels 18 ). Hence, to begin to answer this question, we studied epigenetic development in the hypothalamus during the suckling period (which is relevant potentially to both of our models discussed earlier). Two levels of epigenetic specialisation in the hypothalamus complicate these studies. Within the hypothalamus are specialised regions called hypothalamic nuclei (such as the arcuate nucleus, paraventricular nucleus, etc.) with distinct functions, gene expression patterns, and epigenetic regulation( Reference Flier 40 – Reference Simerly 42 ). Moreover, within the hypothalamus there are many different cell types; the broadest dichotomisation is neurons (which convey information by generating action potentials), and glia (which lack this ability and instead serve broad functions including structural integrity, maintenance, and immune regulation)( Reference Gallo and Deneen 43 ). To ask whether the suckling period is a critical period for cell-type-specific epigenetic development in the mouse, we studied P21 v. P0 hypothalamus, in neuronal and glia DNA separately( Reference Li, Zhang and Baker 44 ). We found dramatic neuron-specific hypermethylation during the suckling period in the mouse. Gene ontology analysis showed that these changes were strongly associated with neurodevelopmental processes. These data indicate that, in the mouse, the suckling period is a critical period for cell-type-specific epigenetic development in the hypothalamus. An obvious next step is to determine if and when similar epigenetic changes occur in the human hypothalamus. Although milestones of brain development are generally thought to occur later in mice than in human subjects, we know very little about how the ontogeny of neuroepigenetics compares between these two species. Such data will be essential to translate findings from mouse studies to man.
Conclusions
Metabolic imprinting of energy balance regulatory mechanisms is likely to be an important factor in developmental programming of obesity. Studying these processes in appropriate inbred animal models currently offers the best opportunity to gain fundamental insights into underlying mechanisms. Importantly, although alteration of epigenetic mechanisms in the central nervous system will almost certainly play a central role, it is imperative to emphasise that this is just one of several potential mechanisms of metabolic imprinting( Reference Waterland and Garza 3 , Reference Waterland 14 ). Gaining a meaningful understanding of how early nutrition affects the establishment of each individuals activity-stat will most likely require integrated analysis of developmental epigenetics in the context of other ontogenic processes that are co-ordinately affected.
Acknowledgements
None.
Financial Support
R. A. W. is supported by the USDA/ARS CRIS 6250-51000-055.
Conflicts of Interest
None.
Authorship
S. Z., J. E., and R. A. W. drafted the manuscript. M. S. B. and G. L. performed studies discussed in the manuscript. All authors edited and approved the final version of the manuscript.
